Hi,

I am using a TC7107A IC to run 3x 7-Segment Common Anode displays (LDS-SMHTA304RISITR). I've drawn up a rough schematic, but it's nothing that deviates much from standard recommended circuits.



The display works as expected... at first. It has no issues on startup, even if I cycle it multiple times per minute or wait 20-40 minutes between startup cycles. BUT, if I go home for the day, the next day when i power up the circuit, several segments no longer work. Several more segments then burn out over the next minute or so. After this, no more segments will burn out. There seems to be a preference for the E segment to burn on all 3 first, followed by G and D on the 100s, sometimes the 10s. If I replace the 7 segment with a new one, it functions perfectly again... until the next day.

Everything is soldered on a custom PCB that was electrical tested, and all the traces and netlists appear to be correct and match the schematic. I drew the schematic manually from the board, just to double check. The resistor between the 5V and the 7 segment anodes is a 25ohm (increased from 5ohm originally), and is there to reduce power dissipation of the displays as per section 9.1 of the IC datasheet. Measured current through that resistor during startup maxes out at about 78mA, which feeds all 22 active segments. This should give approximately 5.1mA per segment, far under the 25mA limit.

I am wondering if leaving the POL(arity) output or AB output unconnected has something to do with it. Can provide more info if needed.

Datasheets available here:
TC7107A: http://ww1.microchip.com/downloads/en/DeviceDoc/21455d.pdf
7-Segment: https://www.lumex.com/spec/LDS-SMHTA304RISITR.pdf

 

Of course after writing this out and replacing all the 7segments at once several segments failed instantly without any form of delay. There must be something wrong with the way the 7 segments are connected to the IC...

 

did you measure segment current? the only reason segments would blow up is high current. where did you get the display from? any chance they are counterfeit?
i see no specific LED current limiting in the chip but output diagram shows that this is smack in the middle of of your displays operating specs.

 

but then again we have not seen the implementation. i hope it is not some horror job, can we see it?

 

Display's were purchased from digikey, so it seems unlikely that they are counterfeit. I wired up a 7 segment independent of the circuit right before I left so I could just make sure that 5mA (or even 20mA) from a lab supply isn't wrecking the segments. then i'll wire the segments off the board via jumper wires, and maybe a 1ohm resistor so i can get some segment current readings on a scope. I can post photos tomorrow of the setup (gone home for the day), but it is a bit tight layout wise so i figured a schematic may be easier to interpret. This has been sitting on my bench for the last few weeks, because I'm frustrated overall with the IC, which began when i found errors in the MQFP pinout.

 

Maybe the output transistors in the chip are shorting somehow.

 

Insert resistors between the IC outputs and each segment inputs to limit current to about 10mA

 

Hey Again,

So more segments burned out upon first power-up this morning.

I took one of the 7 segment digits off the board, wired power to it from the board (Pin 3,8). I then wired the Cathode from the E segment in series with a 10ohm resistor (to measure the current) back to the board. Here is the scope trace from startup. It did not fail, and shows no spikes.



Here is the setup on my bench:


Still frustrated. Any clues as to what could be wrong?

 

I soldered some more segments to the board without the 10ohm resistor. I measured the voltage applied across the LED segment G on startup.

Noise before startup is just a measurement abberation. On startup it is forward biased at 1.87 volts (close to spec of 1.95 typ) for roughly 2 seconds (designed to self test by lighting up all segments for 2 sec). The IC then looks like it applies a reverse bias to the LED of 1V. This appears to fall within the survivable reverse bias of 5V)

Still no failures on the 100s digit that i replaced this morning.

 

Maybe the output transistors in the chip are shorting somehow.

The output transistors appear to be operating correctly, it's a mosfet push pull design. It seems to me that if an output transistor failed there, it would affect only one segment, and replacing the 7 segment digit wouldn't restore it to working condition.

 

Hey Again,

So more segments burned out upon first power-up this morning.

I took one of the 7 segment digits off the board, wired power to it from the board (Pin 3,8). I then wired the Cathode from the E segment in series with a 10ohm resistor (to measure the current) back to the board. Here is the scope trace from startup. It did not fail, and shows no spikes.

View attachment 183553
Here is the setup on my bench:
View attachment 183554

Still frustrated. Any clues as to what could be wrong?

Come, come. THINK! You must have a resistor on every segment, not just on your common anode! The only way you can get away with a single resistor on your common anode is if you never light more than 1 segment at a time. Sorry, but your PCB is bad if it doesn't have the necessary resistors. Do not count on reverse limits, those segments can only handle 100uA.

You need a resistor on each segment, here's the math (per the datasheet):

R = (5-1.95)/0.008
R = 381 Ohms

Use a 360-Ohm resistor on each segment.

 

Come, come. THINK! You must have a resistor on every segment, not just on your common anode! The only way you can get away with a single resistor on your common anode is if you never light more than 1 segment at a time. Sorry, but your PCB is bad if it doesn't have the necessary resistors. Do not count on reverse limits, those segments can only handle 100uA.

You need a resistor on each segment, here's the math (per the datasheet):

R = (5-1.95)/0.008
R = 381 Ohms

Use a 360-Ohm resistor on each segment.

Easy there, cowboy. The TC7107 has constant current segment outputs fixed at 8mA so you don't need (or want) R's in the segment lines. TS's note about the series resistor to the anodes of the display refers to Microchip's advice that the segment current can throw the chip's reference off a bit due to heating. I'd use the diode they recommend if you want to reduce Pd.

The displays are good for 25mA steady current so the 8mA/segment won't be a problem. Vr is 5V so the 1V reverse bias shouldn't be a problem either (assuming you're measuring it correctly).

Measured current through that resistor during startup maxes out at about 78mA, which feeds all 22 active segments. This should give approximately 5.1mA per segment, far under the 25mA limit.

If the display looks right i.e. no segments brighter than the others, I'd tend to think that the design was OK. I've used the 7107 before (way back when it was made by Intersil) with direct connection (no R's) and no problems.

At this point, I'd suspect how you are soldering the parts to the board and if you're doing some mechanical damage. Press on the board to see if any dead segments come back. Check the dead displays with a power supply and resistor (or constant current supply) and see what's still working.

Failing that, I'd contact Lumex tech support and see what they say. Try Microchip, too. I've had two instances over the years when properly implemented displays failed early during burn in. Both problems were related to faulty conductive epoxy internal to the display itself.

Good luck!

 

Easy there, cowboy. The TC7107 has constant current segment outputs fixed at 8mA so you don't need (or want) R's in the segment lines. TS's note about the series resistor to the anodes of the display refers to Microchip's advice that the segment current can throw the chip's reference off a bit due to heating. I'd use the diode they recommend if you want to reduce Pd.

The displays are good for 25mA steady current so the 8mA/segment won't be a problem. Vr is 5V so the 1V reverse bias shouldn't be a problem either (assuming you're measuring it correctly).

If the display looks right i.e. no segments brighter than the others, I'd tend to think that the design was OK. I've used the 7107 before (way back when it was made by Intersil) with direct connection (no R's) and no problems.

At this point, I'd suspect how you are soldering the parts to the board and if you're doing some mechanical damage. Press on the board to see if any dead segments come back. Check the dead displays with a power supply and resistor (or constant current supply) and see what's still working.

Failing that, I'd contact Lumex tech support and see what they say. Try Microchip, too. I've had two instances over the years when properly implemented displays failed early during burn in. Both problems were related to faulty conductive epoxy internal to the display itself.

Good luck!

Thanks, I think you may be on to something with the mechanical damage. These are designed to be reflowed, but I've been hand soldering them. As expected, it was working for as many startups as I could throw at it on Friday, and this morning the failure occurred on the first startup, and failed a little further on the second startup. It is now stable for the day with ALL the segments burned out on the 1s digit, and a single segment on the 10s digit. Oddly, there is no damaged segments in the one I soldered off board, which may be evidence of your point.

 

reflowed the board this time, no failures today. 10+ restarts. Have another set of 7segments on the way from a different manufacturer to test as well.

 

reflowed the board this time, no failures today. 10+ restarts. Have another set of 7segments on the way from a different manufacturer to test as well.

Nice. Let us know how it goes.

 

Not sure how your soldering, but if you had stencils made or at east one for that package it might be easier. Check www.proto-advantage.com if they have single stencils.

if you don't have stencils and can use 63/37 solder, flux and tack the corners. You can then drag solder along the pins with the pins upside down.

You can also try re-flowing in a bed of sand on a hot plate.

the Bismith containing solder pastes are really nice to work with.

 

Nice. Let us know how it goes.

No failures yet! Looks like it could have been hand soldering. This is my prototype run, so i don't have stencil's yet but I will eventually. Basically smeared solder paste as best I could before reflowing. I have a reflow oven, so that part wasn't bad.

I've always used leaded solder paste, interested to try out the bismuth containing stuff now! Any recommended brands?

 

@Domophone Cool!
I don't have any solder recommendations but you could open a thread in General Electronics Chat with that specific question. We have lots of SMT hand-builders here.
Thanks for the feedback.